Blockchain-Based Continuous Knowledge Transfer in Decentralized Edge Computing Architecture

Abstract

Edge computing brings computational ability to network edges to enable low latency based on deploying devices close to the environment where the data is generated. Nevertheless, the limitation of size and energy consumption constrain the scalability and performance of edge device applications such as deep learning, although, cloud computing can be adopted to support high-performance tasks with centralized data collection. However, frequently communicating with a central cloud server brings potential risks to security and privacy issues by exposing data on the Internet. In this paper, we propose a secure continuous knowledge transfer approach to improve knowledge by collaborating with multiple edge devices in the decentralized edge computing architecture without a central server. Using blockchain, the knowledge integrity is maintained in the transfer process by recording the transaction information of each knowledge improvement and synchronizing the blockchain in each edge device. The knowledge is a trained deep-learning model that is derived by learning the local data. Using the local data of each edge device, the model is continuously trained to improve performance. Therefore, each improvement is recorded as the contribution of each edge device immutably in the decentralized edge computing architecture.

1. Introduction

Edge computing leverages computational efficiency to support low latency in the local network where edge devices are deployed to handle the data and environment. The devices are equipped with computing and networking ability to provide services at the network edges. The ability of edge devices enables the devices to access environmental data efficiently [1,2,3]. For responding to the client request, edge computing brings the data and task close to the computing resources that process data in a faster time to reduce the responding delay. The reduced network distance also enables low energy consumption and bandwidth cost by avoiding communicating with the cloud servers through the Internet [4,5,6]. Moreover, keeping data in the network edge to process supports more guaranteed security and privacy by reducing risks on the data communication processes, especially communicating with central cloud servers. Based on the paradigm of edge computing, the computational machines operate applications with a volume of real-time data to provide autonomous and intelligent services in smart spaces including homes, buildings, hospitals, and factories, with smart scenarios [7,8,9,10,11]. Although the applications require sufficient computing ability, cloud services can be adopted to overcome the limitations.

For implementing heavy applications on edge computing to provide services to clients, the tasks and data are offloaded to the cloud server to be processed on the application, such as deep learning. The limitations of devices in the network edge obstruct the scalability and performance of applications due to the size and energy consumption of the hardware for being independent of cloud computing such as central server [12,13,14]. Edge computing is a distributed system that deploys tasks to each edge device. However, operating heavy applications requires tasks to be moved to the cloud, which makes edge computing a part of the centralized system. The important data is transferred to a centralized server which brings security and privacy issues [15,16,17]. For keeping the tasks and data on the network edge, the edge devices distribute the task and data to enable a fully-decentralized system. Knowledge is the extraction of data that can be transferred between edge devices. In the network edge, the edge devices improve knowledge using on-device data without centralizing data to a node based on transferring knowledge, which keeps the data on the devices for security and privacy.

In the process of knowledge transfer between edge devices, blockchain can be adapted to record the transactions based on distributed ledgers. Compared with centralized repositories, the blockchain approach synchronizes the distributed repositories replicated across all devices through community validation in the edge computing network [18,19,20,21]. Blockchain is a ledger that supports decentralized ledger recording by appending a replica of secured transaction information. The blockchain network is a peer-to-peer mesh network where the nodes transfer the content of the ledger by broadcast to others for synchronizing the ledger based on the consensus protocol. The consensus protocol enables the entire blockchain network to update the ledger by validating the common agreement [22,23]. In the validation process, a smart contract is a part of the blockchain node that processes all transactions to control the data by interacting with the blockchain, which performs a consistent logic in all nodes immediately. Bringing blockchain to the network edges as a part of the edge platform enables reliable communications and data integrity. Knowledge transfer is also guaranteed to record consistent transaction information in each edge device.

For continuously improving knowledge in decentralized edge computing, a secure continuous knowledge transfer approach is proposed based on integrating blockchain to the knowledge transfer process. Based on the blockchain network, knowledge integrity is ensured through recording and synchronizing the transaction information of each knowledge improvement in the transfer process between edge devices. The knowledge is comprised of deep learning and local data that are operated in the edge device to improve knowledge continuously. Once an edge device receives the deep learning model, the model is fine-tuned with local data to derive a model to represent the knowledge. In a blockchain node, the smart contract validates the contribution of each improvement based on comparing with previous records. In the implementation of the proposed secure and intelligent edge computing architecture, TensorFlow is used for implementing the deep learning model, Hyperledger Fabric is used for implementing the blockchain network and Flask is used for implementing the web services to expose the functions to the edge computing environment. Based on the proposed decentralized edge computing architecture, each improvement is recorded as the contribution of each edge device immutably for knowledge transfer.

The rest of this paper is structured as follows. In Section 2, the existing edge computing approaches regarding blockchain and knowledge transfer are reviewed, and related solutions of proposed implementation are introduced. Section 3 introduces the proposed decentralized edge computing architecture, including characteristics of knowledge and blockchain in the architecture. Section 4 introduces details of the edge computing platform for presenting the functionality of the knowledge and blockchain agent. Section 5 presents the development details of the proposed secure and intelligent edge computing architecture, including implementation details of the edge computing platform, deployment architecture, and result of the blockchain transaction. Section 6 presents the performances of knowledge improvement over the knowledge transfer process and the contribution of the edge devices. Finally, we conclude this paper and introduce the future direction in Section 7.

2. Related Works

Deploying computing ability close to the network edge enables rich services to be provided around service consumers. Collaborating with cloud computing, edge computing provides heterogeneous and complex services based on delivering collected data to the central server. Intelligence and autonomous applications can also be operated on edge computing without the cloud server with computing ability deployment in constrained devices. For improving the utilization of computing resources, Zhang et al. [24] proposed cooperation of an edge-cloud model to share resources based on a wholesale and buyback business model. Goa et al. [25] proposed a salient object detection approach using distributed cloud-edge architecture for offloading the heavy task to a sufficient cloud server. Wu et al. [26] proposed a video service enhancement strategy based on deploying decision-making in the cloud and caching video data at the edge to improve video service delay. In our previous works [27,28], intelligent tasks were deployed to the high-performance cloud server and microservices providers were deployed edge platform with simple logic to provide efficient services at the network edge. However, data privacy and communication security are risks, due to transferring data from the edge to the cloud.

For operating optimal tasks in the constrained edge computing environment, Li et al. [29] proposed a deep-learning-based tasks assignment approach on distributed edge computing to enable heavy computation in the network edges. Chen et al. [30] proposed an intelligent video surveillance system based on distributing the deep learning models to the multiple edge device to train the model parallelly for improving performance. For keeping the data on the device, federated learning has been proposed to improve the deep learning performance by transferring the model to the central server and aggregating it to derive an updated model [31]. Zheng et al. [32] proposed a distributed hierarchical tensor deep computation model based on deploying low-dimensional tensor spaces to edge devices for reducing the bandwidth consumption and storage requirement of federated learning. Jang et al. [33] proposed a knowledge transfer approach in constrained edge computing architecture by compressing the model and performing knowledge transfer simultaneously. Using the knowledge transfer approach in edge computing, Guo et al. [34] proposed a vehicular network based on sharing the knowledge between nodes to improve object detection accuracy and processing delay.

Blockchain enables the integrity of the transactions in the knowledge transfer in edge computing for reducing the risks in security and privacy, especially transferring data between nodes [35,36]. Swam learning is decentralized federated learning based on the leader-selection mechanism to aggregate the knowledge to one of the nodes in the decentralized network [37]. The approach leverages the blockchain to ensure the integrity of the transactions. However, the leader can be the centralized server due to the selection mechanism base on the computing ability. He et al. [38] applied blockchain to edge computing for security and privacy issues and proposed a machine-learning-based smart contract to allocate the edge computing resources, which improves the efficiency of the blockchain network. Liu et al. [39] proposed a blockchain-based secure data-sharing approach in edge computing using an asynchronous learning method to optimize throughput and energy consumption. Gai et al. [40] proposed the integration of blockchain and edge computing for supporting the validation of smart grids using group signatures and covert channel authorization techniques. Ma et al. [41] proposed a data management approach based on blockchain in edge computing using flexible and configurable methods in consensus and smart contracts.

In a blockchain network, consensus algorithms support reliability to establish trust between participating peers in the decentralized network environment. In the experiment of the proposed system in this paper, the blockchain agent is implemented using the Hyperledger Fabric framework that can be configured with Kafka [42], Raft [43], and Practical Byzantine Fault Tolerance (PBFT) [44], in which the PBFT is configured. The framework is used for the permissioned blockchain network where all participants have a known identity based on the validation mechanism of the membership service provider. For controlling the information of transactions in the blockchain network, chain code performs the role of smart contract and implements the logic in multiple programming languages on the framework. The smart contract is embedded in the blockchain to operate conditional logic based on the permissioned contractual terms automatically without the intervention of network participants [45,46]. Aleksieva et al. [47] proposed a blockchain-based insurance service by handling transactions in a smart contract based on chain code. Javier et al. [48] proposed a decentralized voting system using the smart contract to filter the votes and records on the blockchain. As examples above, smart contract interacts with functions to support heterogeneous application scenarios in decent.

In the proposed decentralized intelligent edge computing, the blockchain is used for providing a trusted record management system in the distributed network. Blockchain technology is applied in many applications to support privacy and security. Due to high-communication-frequency and distributed client peers, federated learning system leverages blockchain to provide high security based on approaches regarding authentication, data sharing, and decentralized storages [49,50,51,52,53]. Nishant et al. [54] leveraged blockchain technology to provide the functionality of storing, validating, and data transfer on a decentralized insurance system securely and reliably. Gong et al. [55] proposed a blockchain-based cyber threat intelligence framework using a standard format to improve the latency of the process. Nguyen et al. [56] proposed a secure and reliable data-sharing approach using blockchain technology for ensuring data integrity in the data process. Zhang et al. [57] proposed a privacy protection scheme for electricity transactions by combining consortium blockchain technology and a continuous double auction mechanism to reduce costs and improve the efficiency of transactions. Mishra et al. [58] integrated blockchain into software-defined networking to provide an efficient approach for defending from distributed denial of service attacks. By comparing with the above works, we integrated blockchain into the fully-decentralized edge computing for recording the contributions while continuous knowledge improvement.

Table 1. Comparison with existing approaches.

Title	Method	Application
He et al. [38]	Machine-Learning-Based Smart Contract	Edge Computing
Liu et al. [39]	Throughput Optimization	Secure Data Sharing
Nguyen et al. [54]	Data Integrity	Secure Data Sharing
Gai et al. [40]	Blockchain-Based Authorization	Smart Grid
Zhang et al. [57]	Transaction Optimization	Smart Grid
Ma et al. [41]	Smart-Contract-Based Configuration	Data Management
Aleksieva et al. [47]	Smart-Contract-Based Transaction Filter	Insurance Service
Javier et al. [48]	Smart-Contract-Based Transaction Filter	Voting System
Yadav et al. [49]	Blockchain-Based Authentication Protocol	Federated Learning
Zhao et al. [50]	Privacy-Preserving Machine Learning	Federated Learning
Li et al. [51]	Distributed Storage using Committee Consensus Protocol	Federated Learning
Peng et al. [52]	Client Peer Authentication	Federated Learning
Hua et al. [53]	Distributed Management Support	Federated Learning
Mishra et al. [58]	Software-Defined-Networking-Based Blockchain	Distributed Denial of Service Attacks
Proposed Approach	Blockchain-Based Continuous Knowledge Transfer	Decentralized Edge Computing

presents the comparisons between the proposed approach and above discussed existing approaches.

3. Proposed Blockchain-Powered Decentralized Edge Computing

The proposed decentralized edge computing is based on knowledge transfer to improve the performance of the intelligence in the network edge without a central server. The knowledge transfer process is secured by the blockchain network to ensure the knowledge integrity for the immutable and trackable contributions of each edge device. An edge device is comprised of computing ability, storage with the local data, and the edge platform. As shown in Figure 1, each edge platform derives the knowledge using the local data and transfers it to the next edge platform that repeats the process using the local data. In the network, the transfer is performed by communicating between edge platforms through web services. Each edge platform provides web services through the web server and accesses the web services of the other edge platform using the web client. An Access Uniform Resource Identifier (URI) is pre-configured in the knowledge transfer scenario. The URIs are used for exposing the web services to receive commands and data that are handled by the functions of the server application.

With the service handlers, the knowledge agent and blockchain agent provide the functionality of intelligence and integrity for operating the proposed decentralized edge computing architecture. The knowledge agent is included in the edge platform to handle the knowledge based on the deep learning model. The knowledge is derived based on local data and deep learning models through fine-tuning in the edge platform. The local data can be Independent Identically Distributed (IID) data that are used for improving knowledge. In decentralized edge computing, over multiple transfers, the knowledge is enabled to provide sufficient performance in the prediction application, then the knowledge can be invoked for prediction services. The blockchain network is operated by the blockchain agent of the edge platform that is triggered by the knowledge operation to validate transactions, and insert blocks into the blockchain. Blockchain is applied to the proposed decentralized edge computing as a distributed ledger to record the information of transfer and contribution in the immutable and trackable mechanism.

The blockchain-based intelligent edge computing is presented in Figure 2 in four layers including layers of edge, intelligence, blockchain, and service. The characteristic of each layer illustrates the functionality of the edge platform in the intelligent and decentralized edge computing architecture.

In the edge layer, edge devices are deployed to interact with the environment through sensors and actuators. The network edge is a constrained environment compared with the Internet. Therefore, the resources are limited to support sufficient computing. Nevertheless, edge computing brings computing ability to the constrained environment to provide services based on data collection and control. Each edge device includes processor, storage, and network equipment that enables the computing to be deployed in the network edge. For collaborating the limited computing ability from multiple edge devices, communication between the devices is required by transferring data through the network. Therefore, the transfer approach overcomes the limitation of the constrained edge environment without invoking the services from the cloud. Also, the environment-collected data are not needed to be offloaded, which reduces the risks to security and privacy. In each edge device, the collected local data is an important property that is used for contributing to knowledge improvement in the knowledge transfer process.

In the intelligent layer, the knowledge is represented by deep learning with the local data of each edge device. Edge devices are connected by the network, where each node transfers the trained deep learning model. The model is comprised of parameters that present knowledge of the data. The data belongs to an edge device that is a platform of an organization for a private purpose. Therefore, the data represent local property that cannot be shared with other entities. Sharing the deep learning model ensures the privacy of data. The first deep learning model is deployed and trained with data. Then, the model is transferred to an edge device and fine-tuned by the local data. In the transfer and fine-tuning process, the performance of the deep learning is changed. Based on the change in performance, the contribution of the edge device is recorded to the ledger in the blockchain system.

In the blockchain layer, each edge platform includes the blockchain agent to enable blockchain ability in the network edge. Blockchain is operated by a distributed and decentralized network where the transactions are recorded by an immutable and trackable mechanism for privacy, security, and transparency. Every transaction is considered to be completely immutable and verified due to the authority of decentralized peers to validate and verify the transactions. The consensus protocol enables all the peers of the blockchain network to achieve a common agreement for accepting a transaction to the ledger. In this way, reliability and trust are established between unknown participants in the blockchain network. Based on the performance change, the smart contract operates the logic to insert the transaction information that is comprised of parameters of transfer and contribution. The transfer parameters illustrate the track of the knowledge for retrieving where the knowledge comes from and where the knowledge is transferred to. The contribution parameters illustrate the improvement of the knowledge in the fine-tuning and the model identifier for accessing the model to apply to services.

In the service layer, the functions of the edge platform are presented by web services and exposed to the network. Some services are exposed to the Internet, such as prediction and retrieving of contributions for further application scenarios. The proposed edge platform is designed by microservices architecture for providing the services that support the functions to be invoked through the web services based on URI. The services of edge platforms are consumed by web clients that are part of the applications for inferencing and developing. Edge devices also request the services from the edge devices for transferring the knowledge to access the URI of the knowledge-receiving service. Based on the exposed services, the functions of knowledge and blockchain agents can be accessed by internal and external entities.

4. Proposed Blockchain-Based Intelligent Edge Computing Platform

The proposed blockchain-powered decentralized edge computing includes multiple edge devices to provide intelligence and data integrity services. The edge device is a physical node in edge computing that provides functionalities based on the edge platform. The proposed edge platform is comprised of device infrastructure, operating platform, knowledge agent, and blockchain agent to provide services in edge computing as shown in Figure 3.

The device infrastructure block enables the edge platform to be operated on edge devices based on sufficient computing ability. Data are also collected by sensors and network modules for storage. For communicating with other edge devices, the network unit can be supported by wired or wireless based on Internet Protocol (IP), because the potential users are distributed on a worldwide scale to improve the knowledge. The operating platform block is comprised of software units including operating system, frameworks, and libraries to support the operation of the proposed functions based on knowledge and blockchain agents.

The knowledge agent includes the deep learning service provider to provide web services that are used for exposing the functions to the Internet as well as the internal use. From the Internet, web clients request services to access the deep learning model and deliver a command to the agent for transferring the model. In both agents, the functions are represented by the services that leverage the microservices development architecture. The deep learning interpreter is used for parsing the transferred model and re-training the model with the local data. The local data is stored in the edge device provided by the participant of the blockchain network. Once a model is transferred to the edge device, the model is updated and stored in the model repository. Then, the model is accessed by web clients to implement prediction services while the model is selected. The model is re-trained in the agent using the fine-tuning approach that updates the parameters of pre-defined deep learning architecture. The operation of deep learning implementation is supported by a deep learning framework such as TensorFlow.

The blockchain agent includes the transaction control service provider, smart contract, ledger repository, and blockchain framework to enable transfer and contribution integrity over the knowledge transfer and contribution. The transaction control service provider provides web services that are used for accessing blockchain-related functions, such as transaction retrieval from internal functions. Smart contract is a program to implement the consistent logic of contractual terms in the blockchain network automatically and immediately. In permissioned blockchain, smart contracts involve the endorsement policy to identify and validate transactions. Once a transaction is entered into the smart contract to meet the condition, the implementation of the contract is operated immediately which is trustful and transparent between participants of the blockchain network. In the peer, a copy of the ledger is stored that presents the blockchain for the blockchain network, In the network, peers are the participants that perform the consensus protocol to validate transactions and add blockchain. The blockchain framework enables the implementation of the modules of the blockchain network participants. For implementing the modules of the blockchain network participants, a blockchain framework can be adopted to build the system such as Hyperledger Fabric.

Figure 4 shows the knowledge transfer scenario in decentralized edge computing based on the blockchain network. First, the initial deep learning model is derived and deployed to edge computing. The model is delivered by the knowledge agent of an edge platform to another edge platform where the knowledge agent receives the model, and parses to tunable parameters to apply on fine-tuning with the local dataset. Once the model is updated by the training step, the agent saves the model for further access by web clients to implement prediction applications. Then, the information of transfer and contribution including identifiers of model, device and previous device, and performance metrics are submitted to smart contract of the blockchain agent. The agent validates the transaction and updates the distributed ledger over all participants of the blockchain network. Finally, the updated model is transferred to the next edge device. In the knowledge transfer process, and each contribution is achieved by the local data and computational support that improves the performance of knowledge.

Figure 5 presents the transaction flow in the blockchain agent for each knowledge transfer. Once the model is updated, the transaction information is submitted to the blockchain agent, and the smart contract invokes the function to pass the transaction information as parameters. The function is named InvokeTx, which is accessed through the command line interface. Inside the InvokeTx function, the QueryAll function is invoked, which is used for retrieving the last inserted contribution to compare. If the new contribution is bigger than the last one, then the distributed ledger is updated to record the new transaction.

5. Development Details

For implementing the proposed blockchain-powered decentralized edge computing platform including knowledge and blockchain agents, we adopted TensorFlow, Flask, and Hyperledger Fabric frameworks to develop the functionality to provide intelligence and data integrity services.

As shown in Figure 6, the implementation is deployed in the Virtual Box platform that emulates the edge device for support computing and networking ability. The edge platform is operated on Ubuntu 22.04 Operating System (OS) where the frameworks and libraries are installed and configured. For enabling the deep learning approach, TensorFlow 2.10 framework is included for providing web services, Flask 2.2.2 framework is included, and for enabling blockchain, Hyperledger Fabric 2.7 is included. Hyperledger Fabric provides a modular architecture that is helpful to develop custom identity management for the permissioned blockchain network.

The knowledge agent is implemented in Python based on a Flask application that exposes functions in web services. Using TensorFlow, the deep learning functions are implemented based on Convolutional Neural Network (CNN) with MNIST dataset. The blockchain agent is deployed based on Docker container where the smart contract is also included to perform the control of transactions. The implementation of smart contract is chain code in Hyperledger Fabric using JavaScript (JS). The chain code is developed by the administrators to handle the transactions in the network for recording the history of all transactions immutably to update states in the ledger. The chain code provides low-level methods including get, put and delete that are used for updating the states of the blockchain. For storing the distributed ledger, the Level Database (DB) is adopted which is performed by a key-value-based database to enable key-based queries. The functions to access the chain code are exposed by the command line interface from the inside of the docker container, and transaction control services are presented by web services.

The implemented edge devices are deployed based on the virtual machines using Virtual Box to emulate computing and network ability. The edge devices are assigned different IP addresses in the edge computing network for configuring the proposed decentralized edge computing to perform the knowledge transfer as shown Figure 7. For distributing the dataset to experiment with IID to analyze the performance, five training datasets and one test dataset are collected from MNIST dataset. The first training dataset is used for deriving the initial CNN model and the model is deployed to the knowledge transfer process to fine-tune with the other four training dataset respectively. The CNN model is comprised of nine layers including the input layer, five convolution-based layers and two classification-based layers, and the output layer.

The blockchain peer is operated in a docker container that is provided by Hyperledger Fabric to control the blockchain network. Each edge device is configured for a different organization with a peer that stores the distributed ledger in the repository. The Hyperledger Fabric node can perform the role of peer, orderer and client that are allowed to be deployed in a node as well as separately. Each peer stores a copy of the distributed ledger that includes the validated transactions based on blockchain. With the smart contract, the peer performs the read and write operations to update the ledger after the consensus validation. The orderer is a necessary entity in a blockchain network that orders the transactions by reaching a consensus. The client is used for accessing the functions of chain code that are invoked by the command line interface and presented by the web services.

Figure 8 presents the implementation result of the transaction that is logged by the blockchain agent by triggering InvokeTx function of chain code. The passed parameters information depicts the model is delivered by an edge device with identifier device2 to this device with identifier device 3. The model identifier is device3:20221207232103 and the prediction accuracy is 89.2%, which is assigned as the contribution of the edge device.

6. Performance Analysis

The proposed knowledge transfer is performed in four edge devices that are emulated by Virtual Box with different network IP addresses. For collecting the performance of the deep learning over the knowledge transfer, 100 times transfers are performed over the edge devices. In the overall transfer process, 22 times are performed to increase the accuracy compared with the previous performance. Therefore, the smart contract records 22 times contributions to the blockchain network. Based on the 22 times improvements, the continuous training in decentralized edge computing improves the performance of the prediction model from less than 20% to higher than 90% using distributed local data of edge devices.

As shown in Figure 9, the prediction accuracy of each fine-tuning is collected, which shows that not all the knowledge improvements are successful to increase accuracy, although fine-tuning tries to improve the performance using more data in the training. Therefore, the implementation of the smart contract acts to record the improvement as the contribution to the ledger.

Figure 10 shows the recorded transaction based on filtering the performance improvements using the smart contract. From the 100 times transfers and fine-tunings, 22 times are performed to improve the prediction accuracy of the deep learning model. The performances are recorded to the ledger as the contribution of edge devices with the transfer information that can be used for developing the prediction applications. The performance is increased from very low accuracy to approximately 90%. The recorded improvements depict that the knowledge transfer in edge computing improves the performance with the decentralized computing ability and data.

Figure 11 presents the total contribution that is collected by summation of improvement for each edge device. The contributions are shared from the total improvement of the knowledge. Based on the result, device 2 improves the accuracy by 28.2% which contributes most, and device 3 improves the accuracy by 8% and provides the least contribution.

7. Conclusions and Future Direction

This paper proposed a blockchain-based continuous knowledge transfer for preventing potential risks on security and privacy issues in decentralized edge computing architecture. A secure continuous knowledge transfer approach is performed for improving knowledge through collaborating with multiple edge devices without a central server. In the proposed system, the blockchain network enables knowledge integrity to be ensured by recording and synchronizing the transaction information of each knowledge improvement in the transfer process between edge devices. The blockchain network is implemented based on Hyperledger Fabric, which is invoked in the blockchain agent to handle the transactions through the smart contract. The smart contract validates the contribution of each knowledge improvement based on comparing with previous records, which provides the immutable and traceable contribution of each edge device. The knowledge is comprised of deep learning and local data that are operated in the edge device to improve knowledge continuously based on fine-tuning. The deep learning approach is implemented using TensorFlow in the knowledge agent to derive the neural network model to enable intelligence in the network edge.

In the future, we will apply a monetary rewarding system to each participant of the blockchain network in the proposed secure and intelligent edge computing architecture. The smart contract performs verification and tracks the contributions of the knowledge improvement. A cryptocurrency framework can be included to provide a rewards delivery system. The monetization framework can facilitate the collaboration of multiple organizations to improve domain knowledge without exposing data to the public.